Investigating intrinsic degradation factors by multi-branch aggregation for real-world underwater image enhancement

Highlights：

• We deeply explore the core difficulty of underwater image enhancement, the complex and diverse degradation factors for underwater images.

• We propose MBANet to simultaneously compensate color and remove veils, to improve the visual quality of underwater images.

• The MBANet establishes a multi-branch multi-variable network to jointly predict three restoration components including the coarse result, veil map, and compensation map, and to perform element-wise division/multiplication among these predicted variables.

• The proposed method outperforms state of the art methods in terms of both execution speed and accuracy.

• We also perform salience object detection on our enhanced images, demonstrating the superiority of our method in complex underwater visual analysis task.

摘要

•We deeply explore the core difficulty of underwater image enhancement, the complex and diverse degradation factors for underwater images.•We propose MBANet to simultaneously compensate color and remove veils, to improve the visual quality of underwater images.•The MBANet establishes a multi-branch multi-variable network to jointly predict three restoration components including the coarse result, veil map, and compensation map, and to perform element-wise division/multiplication among these predicted variables.•The proposed method outperforms state of the art methods in terms of both execution speed and accuracy.•We also perform salience object detection on our enhanced images, demonstrating the superiority of our method in complex underwater visual analysis task.

论文评审过程：Received 21 October 2021, Revised 6 September 2022, Accepted 11 September 2022, Available online 15 September 2022, Version of Record 21 September 2022.